Cooperative haptics for humanoid robot teleoperation

• Autor(a) principal: Barros, João José Oliveira
• Data de Publicação: 2014
• Tipo de documento: Dissertação
• Idioma: eng
• Título da fonte: Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
• Texto Completo: http://hdl.handle.net/10773/14402
• topic: Engenharia mecânica; Robótica; Androides; Aprendizagem automática; Algoritmos de aprendizagem; Dispositivos hápticos

The high degree of complexity associated with the humanoid robotic platforms places some restrictions on the definition of accurate models to the study of motion patterns. Thus, in the context of the University of Aveiro Humanoid Project - PHUA, some work has been done on trying to take advantage of the new concept of robot learning from demonstration, where a human user teleoperates the robot in different motion and balance tasks using a haptic joystick. The data collected during the tests can be used in learning algorithms, so the robot will be able to move, balance and walk on its own. However, the complexity of its kinematic chains, as are the legs in motion and balance operations, makes the learning procedure a lot more challenging and complex due to the shortage of degrees-of-freedom available on the joystick, when compared to those available on the robot. The potential solution lies in the use of two joysticks, one for each leg, in an attempt to ensure a degrees-of-freedom matching between the joystick and the platform. On the other hand, it is also easily recognizable how difficult it could be to teleoperate the real system, not only because of its complexity in terms of degrees-of-freedom but also due to possible errors introduced by mechanical and electronic components. Hence, before trying this operation in the real PHUA platform, this approach is developed and tested in a simulation scenario, making use of an open-source robot simulator, the V-REP. Concomitantly, the haptic interface will generate a force feedback on the haptic device which will be felt by the user, based on the model dynamic response to motion and balance, besides some other disturbances.